Improving CSCL Performance: A Quasiexperimental Study with Control and Intervention Groups Comparing Informative and Suggestive Feedback

Abstract

Computer-supported collaborative learning (CSCL) has been broadly utilized in the field of education. However, learners often face difficulties in improving CSCL performance, including improved knowledge elaboration, knowledge convergence, and coregulation. Therefore, the present study aims to compare the effects of automated informative feedback and knowledge graph-based suggestive feedback on knowledge elaboration, knowledge convergence, and coregulation. This study adopted a convenience sampling method, and a total of 104 undergraduate students registered in a mandatory course voluntarily participated in a quasiexperimental study. The students in experimental Group 1 adopted knowledge graph-based suggestive feedback, the students in experimental Group 2 adopted automated informative feedback, and the students in the control group adopted traditional online collaborative learning without any feedback. The findings revealed that knowledge graph-based suggestive feedback significantly improved group performance, knowledge elaboration, knowledge convergence, and coregulated behaviors compared to informative feedback and traditional online collaborative learning without any feedback. This study has theoretical and practical implications for feedback design and implementation in CSCL practice.

Keywords

collaborative learning knowledge elaboration knowledge convergence informative feedback suggestive feedback

Introduction

Computer-supported collaborative learnaing (CSCL) is a teaching and learning method that centers on how participants learn jointly with all forms of computer technology, such as computers, mobile devices, tablets, and head-mounted displays (Resta & Laferrière, 2007; Stahl et al., 2014). CSCL is an approach that fosters active participation, collaboration, and knowledge sharing among learners. With the growing importance of technology in education, CSCL has become an effective pedagogical approach that is widely used in higher education. One of the key benefits of CSCL is that it promotes group performance, which involves the quality and quantity of products or outcomes that are generated through the collaborative efforts of group members (S. L. Wang & Hong, 2018). By leveraging the power of technology, learners can access shared digital platforms and resources that facilitate communication, coordination, and knowledge sharing among group members, leading to better group performance (Cheng & Chu, 2019). Another key benefit of CSCL is knowledge elaboration, which involves interconnecting, organizing, restructuring, and integrating prior knowledge with new information (Kalyuga, 2009). Through collaborative discussions, peer feedback, and the sharing of resources, learners can engage in knowledge elaboration and build a more comprehensive and sophisticated understanding of the subject matter (Lund, 2019; Mayweg-Paus et al., 2021).

Furthermore, CSCL can promote knowledge convergence, in which the knowledge of individuals learners becomes more similar due to their collaborative efforts (Borge et al., 2018). Through exchanging information, feedback, and ideas, learners can converge on a shared understanding of the subject matter, enhancing group performance by promoting coherence and consistency in their work (Hernández-Sellés et al., 2020). Additionally, CSCL can facilitate coregulation, which involves developing shared norms, goals, and expectations among group members (Ito & Umemoto, 2022). Coregulation promotes a sense of shared responsibility and accountability among group members, leading to better group performance (Zabolotna et al., 2023).

Additionally, CSCL is an approach that promotes active participation, effective collaboration (Meijer et al., 2020), and learning performance improvement (Qureshi et al., 2023). However, Dillenbourg and Fischer (2007) revealed that productive and effective collaborative learning do not occur spontaneously. Learners need timely feedback to facilitate collaborative learning, as it enables learners to track their progress, identify areas for improvement, and adjust accordingly (Hattie & Timperley, 2007). In particular, informative feedback refers to nonsuggestive and noncorrective feedback that provides additional factual information on learning tasks (Deeva et al., 2021). On the other hand, suggestive feedback offers possible strategies or solutions for learners (Deeva et al., 2021). This type of feedback aims to guide the learner toward a specific outcome or solution.

Previous studies have adopted different methods of providing feedback during CSCL. For example, Shin et al. (2020) provided negotiation scaffolding as a kind of suggestive feedback to facilitate problem-solving in the CSCL context. Similarly, Pietarinen et al. (2021) provided teacher guidance as informative feedback during in-person CSCL science classrooms. Additionally, previous studies have examined the impact of feedback on interactions in collaborative language learning (Wen & Song, 2021). Although multiple studies have explored the impact of informative (Tempelaar et al., 2015; Ustun et al., 2023) and suggestive (Jin & Lim, 2019; Sedrakyan et al., 2019) feedback in the classroom context, few studies have compared the impact of informative and suggestive feedback on group performance, knowledge elaboration, knowledge convergence, and coregulation in the CSCL context. By understanding the distinction between these two types of feedback, educators and learners can determine which type is suitable for a particular situation. The study’s results also deepen the understanding of the feedback mechanism. Therefore, it is very important and necessary to carry out the present study to fill these research gaps. The present study aims to compare the effectiveness of suggestive feedback and informative feedback on group performance, knowledge elaboration, knowledge convergence, and coregulation in the CSCL environment. The current study creates real-time knowledge graphs to generate knowledge graph-based suggestive feedback during CSCL activities. A knowledge graph consists of entities and their relationships to represent knowledge (Ji et al., 2021). The following research questions guided this study:

RQ1: Do informative feedback and knowledge graph-based suggestive feedback affect group performance?

RQ2: Do informative feedback and knowledge graph-based suggestive feedback affect knowledge elaboration?

RQ3: Do informative feedback and knowledge graph-based suggestive feedback affect knowledge convergence?

RQ4: Do informative feedback and knowledge graph-based suggestive feedback affect coregulated behaviors?

The current study hypothesizes that knowledge graph-based suggestive feedback has greater impacts on group performance, knowledge elaboration, knowledge convergence, and coregulated behaviors than informative feedback or traditional online collaborative learning.

Literature Review

CSCL

CSCL is a pedagogical approach that leverages technology to support learners in collaboration and coconstruct knowledge (Stahl et al., 2014). With CSCL, learners can communicate and collaborate with their peers, regardless of time and location (Jeong & Hmelo-Silver, 2016). The use of technology tools, such as online discussion forums and collaborative workspaces, enables learners to engage in activities that promote group interaction, knowledge sharing, and reflection. One key aspect of CSCL is its potential to improve group performance (Sun et al., 2025), through knowledge elaboration, knowledge convergence, and coregulation. Below, we provide a more detailed literature review of each aspect.

Group performance is important in evaluating CSCL activities (Zhu & Ergulec, 2023). It refers to the quality and quantity of outcomes generated by peers (Weldon & Weingart, 1993). There are many ways to improve group performance, including providing task-specific collaborative learning experiences (Zambrano R et al., 2023), group awareness support (D. Chen et al., 2024), and group regulation guidance (Yildiz Durak, 2024). However, few studies have been carried out on knowledge graph-based suggestive feedback for boosting group performance.

Knowledge elaboration involves integrating new information with existing knowledge, resulting in a deeper understanding of a topic (Weinstein & Mayer, 1986). In collaborative learning, knowledge elaboration can improve learning outcomes (Lund, 2019) and the quality of collaboration (Stegmann et al., 2012). Knowledge elaboration is crucial for successful collaborative learning (Ding et al., 2011), and studies have shown that it is linked to better learning performance (Denessen et al., 2008). Many researchers have explored ways to facilitate knowledge elaboration through different strategies. For example, Thiel de Gafenco et al. (2024) reported that an information technology-supported cocreation system could enhance knowledge elaboration in vocational education and training. Souza et al. (2024) adopted a concept map to promote knowledge elaboration in the history of science. However, very few studies have utilized knowledge graph-based suggestive feedback to promote knowledge elaboration.

Knowledge convergence refers to increased shared knowledge (Jeong & Chi, 2007), representing a mutual influence among collaborators (Roschelle, 1996). According to Kapur et al. (2011), knowledge convergence emerges from the interaction of individuals, where simplicity at the individual level can result in complexity at the group level. Thus, convergence is a group-level phenomenon that cannot be explicated by individual behavior. Additionally, knowledge convergence is a major goal of collaborative learning (Mercier, 2017) since collaborative learning is a learning activity used to construct common understandings (Roschelle & Teasley, 1995). Knowledge convergence can be achieved through diverse methods, such as collaborative concept mapping (Tan, 2021) and external representation tools (Fischer & Mandl, 2023). However, very few studies have employed knowledge graph-based suggestive feedback to enhance knowledge convergence.

Coregulation is the ability of group members to regulate one another’s learning, which can improve group performance (S. L. Wang & Hong, 2018). By providing feedback and support, group members can help ensure that everyone stays on track and achieves their learning goals. Coregulation is important in productive collaborative learning, and high-performance groups have demonstrated effective coregulated strategies (Su et al., 2023). Previous studies have adopted diverse methods to promote coregulation. For example, Martha et al. (2023) revealed that intelligent agents that act as metacognitive and motivation scaffolding could facilitate coregulation. S.-S. Tseng and Er (2024) suggested that providing feedback and discussion promotes coregulation. Nevertheless, few studies have used knowledge graph-based suggestive feedback to facilitate coregulation.

Informative Feedback

Informative feedback is conceptualized as nonsuggestive and noncorrective feedback that provides additional information on learning tasks (Deeva et al., 2021). Informative feedback emphasizes external postresponse information to inform learning status (Ifenthaler et al., 2021). The purpose of informative feedback is to facilitate learners’ involvement in learning (Ustun et al., 2023). Theobald and Bellhäuser (2022) reported that informative feedback could improve planning strategies and concentration compared with receiving no feedback. Afzaal et al. (2024) revealed that informative feedback facilitates learners’ self-regulation. Information feedback can also facilitate reflection on essays (Gombert et al., 2024). Furthermore, Ustun et al. (2023) reported that learning analytics-based informative feedback could promote self-regulated learning in a flipped learning classroom context. Theobald and Bellhäuser (2022) indicated that informative feedback helped promote self-monitoring and goal achievement. However, the effect of informative feedback is limited since informative feedback cannot provide any suggestions for learners. Therefore, providing suggestive feedback in real time for learners is important.

Suggestive Feedback

Suggestive feedback focuses on advice that helps people improve their ideas by encouraging them to explore, extend, or enhance them (Guasch et al., 2019). It comprises hints and suggestions that are intended to provide guidance on how to move forward and make progress (Deeva et al., 2021). Suggestive feedback provides guidance on improving performance in the future, such as offering instructions for correcting errors (Zhang et al., 2024). The benefits of suggestive feedback have been well documented in previous studies. For example, suggestive feedback positively impacts academic achievements (Jin & Lim, 2019; S. C. Tseng & Tsai, 2007). Tan and Chen (2022) revealed that suggestive feedback could facilitate preservice teachers’ collaborative knowledge improvement. Furthermore, suggestive feedback can significantly improve metacognitive activities, such as planning and monitoring (Guasch et al., 2019). Suggestive feedback is helpful for learning since this form of feedback is more constructive than simply positive or negative opinions (Lai et al., 2020). C. Huang et al. (2023) suggested that suggestive feedback could alleviate learning burnout.

Methodology

Participants

This study included 104 undergraduate students from a famous public university that ranks first among normal universities in China. There were 17 males and 87 females, which was in line with the population distribution of the university. The average age was 18 years (SD = 0.39). The minimum sample size was calculated through G*Power 3.1 software with α = .05 and β = .95 (Faul et al., 2007). The results indicated that the minimum sample size was 84. Therefore, the sample size in this study was statistically fair. This study followed a quasiexperimental design since random assignment was difficult in this study. A quasiexperimental design is more flexible than a true experimental design and can be implemented under natural conditions to reduce threats to validity (Singh, 2021). This study adopted a convenience sampling method, and all participants registered for a mandatory course titled “Fundamentals of Computer Application.” They voluntarily participated in this study and were assigned to 31 groups of three or four students each. The 33 students in experimental Group 1 adopted suggestive feedback on knowledge graphs, the 32 students in experimental Group 2 adopted automated informative feedback, whereas the 39 students in the control group adopted conventional online collaborative learning with no feedback. The participants’ personal information was hidden, and they could quit at any time.

Development of Informative and Suggestive Feedback in the CSCL Platform

This study developed and compared two types of feedback in the CSCL environment: automated informative and automated suggestive feedback. The informative and suggestive feedback was designed by the first author and was confirmed by two domain experts. The automated informative feedback includes a demonstration of the number of online discussions, the duration of the online discussions, the number of interactions with group members, the word cloud of the online discussions, and a summary of the online interactions. The informative feedback can be automatically calculated and demonstrated through the CSCL platform. The online discussion interface of the CSCL platform is shown in Figure 1. Learners can browse the automated informative feedback anytime during their collaborative learning. Figures 2 and 3 show screenshots of informative feedback.

Figure 1.

Screenshot of the CSCL platform.

Figure 2.

Screenshot of informative feedback.

Figure 3.

Screenshot of informative feedback about summary of interaction.

The automated suggestive feedback consists of knowledge graph-based real-time suggestive feedback embedded in the CSCL platform. Implementing knowledge graph-based suggestive feedback includes three steps. The first step is to collect discussion records through the CSCL platform, as shown in Figure 1. The second step is to automatically construct knowledge graphs based on the target knowledge graph and online discussion records. The target knowledge graph denotes the correct target knowledge and relationships. Finally, constructing knowledge graphs includes entity recognition and relation extraction.

This study integrates sequence tagging and keyword matching to recognize entities automatically. Sequence tagging is achieved through a deep neural network (DNN) model named BERT-BiLSTM-CRF, which was validated by Zheng et al. (2023). The main reason for using the BERT-BiLSTM-CRF model was that, compared with the other competing models, this model achieves the highest performance, as shown in Table 1. The relation extraction is conducted by querying the target knowledge graph. Therefore, our system can automatically construct and demonstrate the activated and inactivated knowledge graphs of each group. The activated public knowledge graph can also be automatically demonstrated. The activated knowledge graph consisted of knowledge nodes and their relationships activated by one group member. The activated public knowledge graph consisted of knowledge nodes and their relationships activated by all group members. The third step automatically provides suggestive feedback according to the knowledge graphs. The suggestive feedback includes textual feedback and recommended learning resources. The learning resources are recommended on the basis of the target knowledge graph. Table 2 shows the conditions and content of the suggestive textual feedback according to the constructed knowledge graphs. These feedback conditions and content are set and tested on the basis of a pilot study conducted by the authors’ team. Figures 4 and 5 show screenshots of suggestive feedback based on an activated knowledge graph and activated public knowledge graphs. The number beside each knowledge node in the knowledge graph represents the activation quantity that can be computed on the basis of the algorithm validated by the Zheng (2017).

Table 1.

The Performance of Different Models for Entity Recognition.

Models	Accuracy	Weighted-precision	Weighted-recall	Weighted-F1
BERT	0.85	0.58	0.64	0.57
BERT-LSTM-CRF	0.92	0.90	0.92	0.91
BERT-BiLSTM-CRF	0.93	0.93	0.93	0.92

Table 2.

The Suggestive Feedback Based on Knowledge Graphs.

Conditions	Suggestive feedback
The number of activated knowledge nodes is less than one-third of the nodes in the target knowledge graph	It was found that only a few knowledge nodes have been activated thus far. It is advised that the group leader organizes all members to revisit the collaborative learning tasks and discuss the inactivated knowledge. Please conduct in-depth discussion to establish relationships between the activated knowledge and inactivated knowledge. Please link prior knowledge and new information to promote knowledge elaboration.
The number of activated knowledge nodes is between one-third and two-thirds in the target knowledge graph	It was found that your group made progress and more than one-third of the nodes were activated. It is suggested that your group needs to continually discuss the inactivated knowledge nodes. You can propose new ideas and continually incorporate previous knowledge and new information. Come on!
The number of activated knowledge nodes is greater than two-thirds of the nodes in the target knowledge graph	It was found that your group has activated most of the knowledge nodes in the target knowledge graph. It is suggested that your group check and fill in the gaps based on the inactivated knowledge nodes. Please also discuss the rest of the inactivated knowledge nodes and complete further collaborative learning. Your group can also continue to further discuss and elaborate on the activated knowledge nodes. Come on!
The number of activated nodes in the group public knowledge graph is fewer than one-third of the nodes in the activated knowledge graph	It was found that the knowledge convergence level of your group is very low. It is suggested that each group member browse the activated public knowledge graph to detect differences. Please focus on collaborative learning tasks and complete them together. Do not go off-topic.
The number of activated nodes in the public knowledge graph is between one-third and two-thirds in the activated knowledge graph	It was found that the knowledge convergence level of your group made progress, but still needs to be improved further. It is suggested that your group continually discusses the knowledge nodes that has not appeared in the public knowledge graph. The group leader should organize everyone to concentrate on the collaborative learning task and achieve a common understanding.
The number of activated nodes in the public knowledge graph is greater than two-thirds of nodes in the activated knowledge graph	Although the knowledge convergence level of your group is high, it can be improved further. It is suggested that your group expand the number of knowledge nodes in the public knowledge graph. The group leader should organize everyone to conduct reflection and the evaluation of solutions to achieve agreement. Come on!

Figure 4.

Screenshot of suggestive feedback based on an activated knowledge graph.

Figure 5.

Screenshot of suggestive feedback based on a public knowledge graph.

Experimental Procedure

This experiment included five phases: pretesting, providing basic knowledge and skills about the computer network, conducting online collaborative learning, submitting group products, and interviewing. The first phase was to complete the pretest for 20 min to examine the participants’ prior knowledge of the computer network to ensure there was no difference in the previous knowledge between the control and experimental groups. The second phase involved instructing basic knowledge and skills about computer networks and how to utilize the online collaborative learning platform. In the third phase, the participants in two experimental groups and one control group engaged in online collaborative learning for 6 weeks. The online collaborative learning task involves solving problems related to how to build a local area and wireless network; troubleshooting network failures; and retrieving, processing, and utilizing relevant information. The tasks and durations were identical for all the participants. The main difference was that the participants in experimental Group 1 were provided with suggestive feedback based on knowledge graphs, the participants in experimental Group 2 were given information feedback, and the students in the control group did not receive any feedback. After 6 weeks, all the groups submitted the solutions as group products. The last phase involved conducting semistructured interviews with the participants in experimental Group 1 online for 30 min.

Instruments

The instruments used in the present study included a pretest and interview protocol. The purpose of the pretest was to examine the participants’ prior knowledge of computer networks. It consists of 20 multiple-choice items and 5 short-answer items. Each multiple-choice item was scored 3, and each short-answer item was scored 8. The total score was 100. The pretest was designed by a knowledgeable teacher who taught computer networking for 10 years. The validity of the pretest was examined and confirmed by two domain experts. The interrater reliability of the pretest calculated through the kappa value was 0.72. The interview protocol included interview questions. All the interview questions were designed by the first author and confirmed by two experts. The sample questions included the following: Do you think knowledge graph-based suggestive feedback can improve group performance? If so, why? Do you think knowledge graph-based suggestive feedback contributes to promoting knowledge elaboration? If so, why? Do you think the knowledge graph-based suggestive feedback can facilitate knowledge convergence? If so, why? Do you think the knowledge graph-based suggestive feedback can promote coregulated behaviors? If so, why?

Data Collection and Analysis Methods

The data collection methods used in this study include tests, online discussions, rating group products, and interviews. The data analysis method in this study included a knowledge graph analysis method, a lag sequential analysis method, a content analysis method, and a statistical analysis method. First, a content analysis method was used to analyze the group products. Group products are solutions to problems related to how to build networks, troubleshoot network failures, and process information. The group products were analyzed and evaluated according to predefined criteria that were adapted from previous studies (Zheng et al., 2023; Hwang & Chang, 2021), as shown in Table 3. Two experts highly experienced in computer networking evaluated all the group products, and the kappa value of the interrater agreement was 0.81. Group performance was quantified through the scores of group products.

Table 3.

Assessment Criteria for Group Products.

Dimensions/scores	16–20	15–11	6–10	1–5
Precision (20)	Precise solutions, examples, and explanations.	Precise solutions and explanations, but incorrect examples.	Precise solutions, but improper explanations and instances.	Wrong solutions, examples, and explanations.
Rationality (20)	The solutions, examples, and explanations were reasonable.	The solutions, explanations, and examples were partly reasonable.	Only the solutions were reasonable.	The solutions, explanations, and examples were not reasonable.
Feasibility (20)	All the solutions were feasible.	The plans were partly possible.	The plans were not possible.	The plans were lacking.
Originality (20)	The group products included many original solutions.	The group products included at least one new solution.	The group products included several hackneyed and unoriginal solutions.	All the solutions were hackneyed and unoriginal.
Completeness (20)	The plans were well-organized and complete.	The plans were basically complete and achieved the requirements.	Some of the plans were disorganized and incomplete.	All the plans were very messy and incomplete.

Second, knowledge elaboration and knowledge convergence were analyzed using the knowledge graph analysis method. This analysis method includes drawing a target knowledge graph, segmenting group discussion records, and calculating knowledge elaboration and knowledge convergence levels (Zheng, 2017). Knowledge elaboration was calculated through the weighted path length of the activated spanning tree (Zheng, 2017). Knowledge convergence was equal to the activation quantity of the public knowledge graph (Zheng, 2017). Two coders analyzed all group discussion records using a developed tool. The interrater reliability calculated by Kappa values for knowledge elaboration and knowledge convergence reached 0.9 and 0.91, respectively.

The coregulated behaviors were analyzed through the lag sequence analysis method. Table 4 shows the coding scheme of coregulated behaviors, which was adapted from Kielstra et al. (2022) and Zheng et al. (2023). Two coders analyzed all the group discussion records independently according to the coding scheme. The interrater agreement calculated by the Kappa value reached 0.91. The discrepancies between the raters were discussed and resolved face-to-face. GSEQ 5.1 software (Quera et al., 2007) was subsequently used to analyze coregulated behavioral patterns.

Table 4.

The Coding Scheme for Coregulated Behaviors.

Dimensions	Examples
Orienting goals (OG)	“We need to orient the learning goals of our group first.”
Making plans (MP)	“Let’s discuss how to make reasonable plans.”
Enacting strategies (ES)	“The current strategies are not appropriate for troubleshooting network failures. We can find another solution.”
Monitoring and controlling (MC)	“Time flies. We have only completed one subtask and need to speed up.”
Evaluating and reflecting (ER)	“We need to reflect on and conduct an assessment of our group product first; then, we can modify and refine it further.”
Adapting metacognition (AM)	“We have to adjust our plans based on the analysis results.”

Additionally, semistructured interviews were performed to obtain perceptions of knowledge graph-based suggestive feedback. All the interview records were independently analyzed by two coders based on thematic analysis methods (Braun & Clarke, 2006). The interview records were divided into four themes: improving group performance, promoting knowledge elaboration, promoting knowledge convergence, and facilitating coregulated behaviors. The interrater reliability of the interviews was 0.9, indicating high reliability.

Results

Impacts on Group Performance

We employed one-way ANCOVA to compare the impacts of the two types of feedback on group performance. Before performing ANCOVA, all datasets were examined for normality distribution through the Kolmogorov–Smirnov test. The findings revealed that all datasets had a normal distribution (p > .05). Furthermore, the homogeneity of variance for group performance was not violated (F = 1.125, p = .339). The results of the homogeneity test of the regression coefficient indicated that the homogeneity of the regression was not violated (F = 0.303, p = .741). Hence, ANCOVA could be performed using the feedback type as the independent variable, group performance as the dependent variable, and the pretest score as the covariant variable. Table 5 shows the ANCOVA results, and the findings revealed there was a significant difference in group performance among the three groups (F = 64.30, p < .001). Moreover, the effect size calculated through the omega squared (Lakens, 2013) indicated there was a large effect size (ω² = 0.83) according to Cohen (1988). Furthermore, post hoc analysis was carried out using the least significant difference (LSD) test due to its ability to provide more precise results than other tests. The findings revealed that the group performance of experimental Group 1 was significantly greater than those of experimental Group 2 and the control group. The interview results also indicated that knowledge graph-based suggestive feedback was very helpful because their groups often refined group products according to the suggestive feedback (100%). A total of 91% of the interviewees stated that knowledge graph-based suggestive feedback contributed to the acquisition of new knowledge and skills. Therefore, the learners who adopted the knowledge graph-based suggestive feedback approach had better group performance than those who employed either the informative feedback approach or the traditional online collaborative learning approach.

Table 5.

ANCOVA of Results in Group Products.

Group	N	Mean	SD	Adjusted mean	SE	F	ω²	Post hoc
(1) Experimental group 1	33	90.63	2.37	90.87	1.80	64.30***	0.83	(1) > (2)
(2) Experimental group 2	32	75.60	3.37	75.77	1.45			(1) > (3)
(3) Control group	39	72.00	2.70	71.56	3.01

***

p < .001.

Impacts on Knowledge Elaboration

This study used ANCOVA to analyze the impacts of two types of feedback on knowledge elaboration. All datasets were normally distributed according to the Kolmogorov–Smirnov test (p > .05). Furthermore, neither the homogeneity of variance (F = 1.573, p = .225) for knowledge elaboration nor the homogeneity of regression (F = 0.154, p = .858) were violated. Therefore, ANCOVA could be performed. Table 6 presents the results. A significant difference in knowledge elaboration was found among the three groups (F = 17.15, p < .001). The omega squared (ω²) was 0.56, implying a large effect size (Cohen, 1988). Moreover, post hoc analysis was conducted using the LSD test. The results revealed that the knowledge elaboration of experimental Group 1 was significantly greater than experimental Group 2 and the control group. Additionally, all the interviewees stated that knowledge graph-based suggestive feedback helped them link prior knowledge with new information (100%). A total of 91% of the interviewees stated that knowledge graph-based suggestive feedback stimulated them to generate new ideas. Therefore, the learners who adopted the knowledge graph-based suggestive feedback approach had greater knowledge elaboration than those who employed either the informative feedback approach or the traditional online collaborative learning approach.

Table 6.

ANCOVA of Results in Knowledge Elaboration.

Group	N	Mean	SD	Adjusted mean	SE	F	ω²	Post hoc
(1) Experimental group 1	33	568.02	151.87	583.90	83.49	17.15***	0.56	(1) > (2)
(2) Experimental group 2	32	210.31	140.80	221.74	67.49			(1) > (3)
(3) Control group	39	85.96	90.25	57.06	139.65

***

p < .001.

Impacts on Knowledge Convergence

The Kolmogorov–Smirnov test findings demonstrated that all datasets satisfied a normal distribution (p > .05). After confirming the hypothesis of homogeneity of variance (F = 0.628, p = .439) and homogeneity of regression (F = 0.472, p = .629), ANCOVA was conducted. Table 7 shows the results. A significant difference in knowledge convergence was revealed among the three groups (F = 4.78, p = .01). The omega squared (ω²) value was 0.26, implying a large effect size (Cohen, 1988). Additionally, post hoc analysis was conducted via the LSD test. The results revealed that the knowledge convergence of experimental Group 1 was significantly greater than that of experimental Group 2 and the control group. Moreover, all the interviewees told us that the knowledge graph-based suggestive feedback reminded them not to be off-topic and to concentrate on the learning task. A total of 91% of the interviewees stated that the activated public knowledge graph and suggestive feedback promoted them to achieve shared understanding. Therefore, the learners who adopted the knowledge graph-based suggestive feedback approach had greater knowledge convergence than those who employed either the informative feedback approach or the traditional online collaborative learning approach.

Table 7.

ANCOVA of Results in Knowledge Convergence.

Group	N	Mean	SD	Adjusted mean	SE	F	ω²	Post hoc
(1) Experimental group 1	33	191.28	116.14	196.43	50.15	4.78*	0.18	(1) > (2)
(2) Experimental group 2	35	77.95	65.06	81.66	40.54			(1) > (3)
(3) Control group	39	9.74	9.63	.368	83.88

p < .05.

Impacts on Coregulated Behavioral Patterns

Tables 8 –10 present the adjusted residuals of the two experimental groups and one control group, respectively. If the adjusted residuals are greater than 1.96, the behavior sequence is significant (Bakeman & Quera, 2011). The findings revealed nine significant coregulated behavioral transitions in experimental Group 1, including OG → OG (orienting goals continually), OG → MP (making plans afterward setting goals), MP → MP (making plans continually), MP → MC (monitoring and controlling after making plans), ES → ES (enacting strategies continually), ES → AM (adapting metacognition after enacting strategies), MC → MC (monitoring and controlling continually), ER → ER (evaluating and reflecting continually), and AM → MC (monitoring and controlling after adapting metacognition).

Table 8.

Adjusted Residuals of the Experimental Group 1.

Starting behavior	Subsequent behavior
	OG	MP	ES	MC	ER	AM
Orientating goals (OG)	8.59*	7.05*	−4.59	−3.88	−2.42	0.45
Making plans (MP)	−2.02	10.75*	−7.19	2.02*	−2.65	−1.81
Enacting strategies (ES)	−3.78	−9.4	19.14*	−9.83	−0.19	2.15*
Monitoring and controlling (MC)	−0.14	−0.7	−8.94	9.68*	−1.04	−1.23
Evaluating and reflecting (ER)	1.08	−3.04	−2.41	−0.73	9.26*	−0.27
Adapting metacognition (AM)	0.54	−0.44	−2.52	2.58*	−1.17	1.94

p < .05.

Table 9.

Adjusted Residuals of the Experimental Group 2.

Starting behavior	Subsequent behavior
	OG	MP	ES	MC	ER	AM
Orientating goals (OG)	7.85*	1.62	−4.13	−0.2	−0.28	−0.48
Making plans (MP)	−0.81	14.17*	−6.59	−1.25	−2.09	−0.76
Enacting strategies (ES)	−2.79	−6.02	13.11*	−7.53	1.8	−0.44
Monitoring and controlling (MC)	0.03	−3.35	−6.39	8.1*	−2.03	1.17
Evaluating and reflecting (ER)	−0.25	−0.26	0.4	−1.88	5.25*	−0.43
Adapting metacognition (AM)	−0.48	−0.76	1.46	−0.49	−0.43	−0.13

p < .05.

Table 10.

Adjusted Residuals of the Control Group.

Starting behavior	Subsequent behavior
	OG	MP	ES	MC	ER	AM
Orientating goals (OG)	15.39*	−0.92	−2.26	−3.97	−4.66	−0.83
Making plans (MP)	−2.48	17.18*	−2.90	−5.98	−7.80	−0.38
Enacting strategies (ES)	−2.12	−5.30	7.45*	0.57	0.33	1.79
Monitoring and controlling (MC)	−4.11	−6.22	−0.52	9.41*	0.50	0.87
Evaluating and reflecting (ER)	−3.25	−7.11	−0.71	−0.50	11.11*	−1.23
Adapting metacognition (AM)	−0.81	0.64	0.46	−0.17	−0.22	−0.18

p < .05.

Additionally, five significant behavioral sequences appeared in experimental Group 2, namely, OG → OG (orienting goals repeatedly), MP → MP (making plans repetitively), ES → ES (enacting strategies repetitively), MC → MC (monitoring and controlling repetitively), and ER → ER (evaluating and reflecting repetitively). There was no metacognition adaptation in experimental Group 2, which also indicated that the participants in experimental Group 2 could not coregulate in collaborative learning.

Moreover, only five significant behavioral sequences appeared in the control group, including OG → OG, MP → MP, ES → ES, MC → MC, and ER → ER. There were no coregulated behavior transitions in the control group. Hence, the participants in the control group could not coregulate with peers during collaborative learning. Figure 6 shows the behavioral transition diagrams of the two experimental groups and one control group.

Figure 6.

Behavioral transition diagrams of the two experimental groups and one control group.

Table 11 shows the significant behavior sequences that occurred only in experimental Group 1. Monitoring and controlling, as well as adapting to metacognition, were the most crucial coregulated behaviors. Moreover, there were two major differences in coregulated behavioral patterns among the three groups. First, the amount of significant coregulated behavior of experimental Group 1 was greater than that of experimental Group 2 and the control group. Therefore, the participants in experimental Group 1 could coregulate themselves using the suggestive feedback. Second, experimental Group 1 had more behavioral transitions than the experimental Group 2 and the control group. The main reason for these results is that experimental Group 2 received only informative feedback and that the control group did not receive any feedback. Additionally, all the interviewees believed that the knowledge graph-based suggestive feedback motivated them to coregulate each other, monitor the online collaborative learning progress, reflect and evaluate collaborative learning processes and group products, and adapt goals, plans, and strategies. Therefore, the learners who adopted the knowledge graph-based suggestive feedback approach demonstrated more coregulated behaviors than those who employed either the informative feedback approach or the traditional online collaborative learning approach.

Table 11.

Significant Behavior Sequences That Only Occurred in the Experimental Group 1.

Starting behavior	Subsequent behavior
	OG	MP	ES	MC	ER	AM
Orientating goals (OG)	OG → MP
Making plans (MP)	MP → MC
Enacting strategies (ES)			ES → AM
Monitoring and controlling (MC)
Evaluating and reflecting (ER)
Adapting metacognition (AM)			AM → MC

Discussion

Discussion of Major Findings

This study revealed that knowledge graph-based suggestive feedback had more significant and positive impacts on group performance than informative feedback. There are two possible reasons for these results. First, knowledge graph-based suggestive feedback facilitates the coconstruction of knowledge and skills through activated, inactivated, and public knowledge graphs and suggestive feedback, which improves group performance. Knowledge graph-based suggestive feedback also promoted the acquisition of new knowledge and skills. These results also corroborated the findings of Cui and Yu (2019), who reported that using knowledge graphs results in better performance. Second, the knowledge graph-based suggestive feedback provided immediate and positive feedback, as well as recommended learning resources, which helped learners improve group performance. These outcomes are in line with those of C. M. Chen et al. (2023), who reported that positive and constructive feedback improves learning performance. The findings were also confirmed by most interviewees, who said, “Our group believed that the knowledge graph-based suggestive feedback approach is amazing, and the functionalities are very powerful. It can automatically construct knowledge graphs and provide suggestive feedback immediately based on our discussion records, which contributes to improving group performance. We truly like it.”

This study revealed that knowledge graph-based suggestive feedback had significant and positive impacts on knowledge elaboration. There are two possible reasons for this result. First, knowledge graph-based suggestive feedback encouraged learners to activate more knowledge and to link prior knowledge with new information, which facilitated knowledge elaboration. Weinstein and Mayer (1986) proposed that knowledge elaboration could be achieved by incorporating new information with prior knowledge. Similarly, Erkens et al. (2019) concluded that obtaining specific information about partners and learning content linked with prior knowledge guides metacognitive regulation, resulting in deeper knowledge elaboration. Second, knowledge graph-based suggestive feedback also promoted the further elaboration of knowledge. These results also corroborated the statements of Thiel de Gafenco et al. (2024), who revealed that feedback could facilitate knowledge elaboration. This finding is consistent with that of Hsieh and Tsai (2012), who also reported that providing feedback stimulates learners to elaborate on knowledge. This result was verified by many interviewees, who stated that “The automatically constructed knowledge graphs guide us to link the activated with inactivated knowledge to elaborate knowledge further. In addition, suggestive feedback is very useful for facilitating the integration of previous knowledge with new information. We have learned a lot from the knowledge graph-based suggestive feedback.”

This study revealed that knowledge graph-based suggestive feedback had significant and positive impacts on knowledge convergence. There are three possible reasons. First, knowledge graph-based suggestive feedback contributes to achieving a common understanding of collaborative learning tasks, which significantly promotes knowledge convergence. These results are also consistent with those of the study by Sadita et al. (2020), who revealed that a collaborative map can serve as an artifact that represents shared understanding and group consensus. Second, knowledge graph-based suggestive feedback serves as a group awareness tool, which facilitates knowledge convergence. Group awareness tools display awareness information about collaborative learning processes, which contributes to achieving a shared understanding and convergence of knowledge (Kwon, 2020). Third, knowledge graph-based suggestive feedback stimulates group members to achieve agreement, improving knowledge convergence. This result is consistent with that of W. Chen (2017), who revealed that constructive feedback contributes to achieving knowledge convergence. In contrast, the control group only received informative feedback without any suggestions. These factors could explain why knowledge graph-based suggestive feedback significantly impacts knowledge convergence.

This study revealed that knowledge graph-based suggestive feedback significantly promoted coregulation. The main reason for this finding was that the knowledge graph-based suggestive feedback demonstrated activated and inactivated public knowledge graphs, as well as suggestive feedback, facilitating coregulation among group members. For example, group members could monitor and reflect online collaborative learning processes by browsing knowledge graphs and suggestive feedback. As many interviewees mentioned, “Our group members could co-regulate behaviors according to the suggestive feedback. When we found that there were few knowledge nodes in the public knowledge graph, we focused on collaborative learning tasks and converged our ideas further.” Additionally, suggestive feedback was provided for learners according to the analysis results of knowledge graphs in real time, which also promoted coregulation among group members. These results also corroborated the findings of Järvelä and Hadwin (2013), who determined that constructive feedback could promote coregulation in CSCLs. This finding was also supported by many interviewees who stated: “The suggestive feedback provided the concrete advice, which prompted us to co-regulate each other further. For example, when the off-topic information decreased significantly, we co-regulated our behaviors to concentrate on collaborative learning tasks immediately based on suggestive feedback.”

Implications

This study has several practical and technical implications for teachers, practitioners, and developers. First, knowledge graph-based suggestive feedback is useful for improving group performance, knowledge elaboration, knowledge convergence, and coregulated behaviors. The large effect size of the knowledge graph-based suggestive feedback also revealed its practical significance in improving group performance. Research has indicated that incorporating suggestive feedback enhances students’ performance as they actively participate in cognitive and metacognitive activities (Guasch et al., 2019; Tan & Chen, 2022). Furthermore, as a type of computer-mediated feedback, knowledge graph-based suggestive feedback can greatly reduce teachers’ workload. Bahari (2021) also reported many benefits and potentials of computer-mediated feedback, such as improving learning achievements and reducing workload. Therefore, adopting knowledge graph-based suggestive feedback in CSCL practice is recommended.

Second, this study revealed that the power of knowledge graph-based suggestive feedback is stronger and more effective than that of informative feedback. This finding has important implications for designing and implementing feedback in CSCL practice. When designing feedback, suggestive rather than informative feedback is recommended. Researchers have reported a stronger association between suggestive feedback and higher-order thinking skills (Jiang et al., 2023; Shao et al., 2024). Specific and elaborate suggestions can trigger deeper thinking and cognitive conflict among students, causing them to question their beliefs and argue about, analyze, and experiment with new ideas, thereby fostering critical thinking skills (Barahona et al., 2023; Jiang et al., 2023; Kerman et al., 2024). Suggestive feedback can provide direction on progressing and improving what learners have done, which contributes to boosting learning performance (Sedrakyan et al., 2020). Suggestive feedback contributes to more cognitive and metacognitive activities (Guasch et al., 2019). Furthermore, suggestive feedback can engage learners in acting since feedback should be considered acting rather than telling (Espasa et al., 2018). Suggestive feedback also provides useful guidance on where to proceed (Latifi et al., 2021). Suggestive feedback can also feed forward because only suggestive feedback can influence recipients’ actions (Tan & Chen, 2022).

Third, this study adopted deep neural network models to generate knowledge graph-based suggestive feedback. As a type of artificial intelligence technology, deep neural network models are very promising for entity recognition and the construction of knowledge graphs. Deeva et al. (2021) proposed that automated feedback can be provided through artificial intelligence or other advanced learning technologies. Artificial intelligence technology-mediated feedback design and implementation have great potential for triggering engagement and enhancing problem-solving (Z. Huang et al., 2024). Therefore, when designing feedback, using deep neural network models to construct knowledge graphs and provide suggestive feedback is recommended. Furthermore, innovative deep neural network models should be developed to improve accuracy and performance in the future.

Limitations

The present study has several limitations. First, only 104 college students participated in the present study due to the COVID-19 pandemic. Future research should increase the sample size to investigate the two types of feedback further. Second, this study engaged participants in completing one collaborative learning task in one learning domain. Therefore, caution should be taken when generalizing the findings. Future research should compare the two types of feedback through different learning tasks. Third, this study only examined the impacts of the two types of feedback on group performance, knowledge convergence, knowledge elaboration, and coregulated behaviors. Future research should also investigate the impacts of these two types of feedback on learning engagement, high-order thinking skills, and other variables.

Conclusions

This study proposed and validated the idea that knowledge graph-based suggestive feedback has more significant and positive impacts on group performance, knowledge elaboration, knowledge convergence, and coregulated behaviors than informative feedback. The main contribution of the present study is threefold. First, it proposes and compares the effects of informative feedback and knowledge graph-based suggestive feedback in CSCL settings. The second contribution is finding suggestive feedback has a more substantial impact on collaborative learning performance than informative feedback. Third, it enriches the understanding of different types of feedback and increases the knowledge of the CSCL area. The findings of the current study can be applied to provide feedback and improve learning performance in the CSCL context. This study also paves the way for enriching the theory and practice of feedback for the research community.

Footnotes

ORCID iD

Lanqin Zheng

Ethical Considerations

All participants in this study agreed to participate in this study by giving verbal and written consent. Furthermore, all ethical guidelines were kept.

Author Contributions

L.Z., K.K.B.: conceptualization. L.Z., K.K.B., M.L., N.K.J.: methodology, formal analysis, investigation, data collection, data curation, writing original draft. L.Z., K.K.B.: writing—review and editing; L.Z.: funding acquisition. All authors read and approved the final manuscript.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study is funded by the Key Technologies for Personalized Learning Driven by Educational Big Data and the Demonstration Application (2023YFC3341200).

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The datasets used during the current study are available from the corresponding author on reasonable request.

References

Afzaal

Zia

Nouri

Fors

(2024). Informative feedback and explainable AI-based recommendations to support students’ self-regulation. Technology, Knowledge and Learning, 29(1), 331–354.

Bahari

(2021). Computer-mediated feedback for L2 learners: Challenges versus affordances. Journal of Computer Assisted Learning, 37(1), 24–38.

Bakeman

Quera

(2011). Sequential analysis and observational methods for the behavioral sciences. Cambridge University Press.

Barahona

Nussbaum

Martin

Meneses

Arriagada

Di Serio

Hilliger

(2023). Technology-scaffolded peer assessment for developing critical thinking in pre-service teacher training: The importance of giving feedback. Educational Technology Research and Development, 71(2), 667–688.

Borge

Ong

Y. S.

Rosé

C. P.

(2018). Learning to monitor and regulate collective thinking processes. International Journal of Computer-Supported Collaborative Learning, 13, 61–92.

Braun

Clarke

(2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101.

Chen

C. M.

M. C.

Huang

Y. L.

(2023). Developing an instant semantic analysis and feedback system to facilitate learning performance of online discussion. Interactive Learning Environments, 31(3), 1402–1420.

Chen

Zhang

Luo

Zhu

Lin

(2024). Effects of group awareness support in CSCL on students’ learning performance: A three-level meta-analysis. International Journal of Computer-Supported Collaborative Learning, 19(1), 97–129.

Chen

(2017). Knowledge convergence among pre-service mathematics teachers through online reciprocal peer feedback. Knowledge Management & E-Learning: International Journal, 9(1), 1–18.

10.

Cheng

L. C.

Chu

H. C.

(2019). An innovative consensus map-embedded collaborative learning system for ER diagram learning: Sequential analysis of students’ learning achievements. Interactive Learning Environments, 27(3), 410–425.

11.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Routledge.

12.

Cui

(2019). Fostering deeper learning in a flipped classroom: Effects of knowledge graphs versus concept maps. British Journal of Educational Technology, 50(5), 2308–2328.

13.

Deeva

Bogdanova

Serral

Snoeck

De Weerdt

(2021). A review of automated feedback systems for learners: Classification framework, challenges and opportunities. Computers & Education, 162, 104094.

14.

Denessen

Veenman

Dobbelsteen

Van Schilt

(2008). Dyad composition effects on cognitive elaboration and student achievement. The Journal of Experimental Education, 76(4), 363–386.

15.

Dillenbourg

Fischer

(2007). Computer-supported collaborative learning: The basics. Zeitschrift für Berufs- und Wirtschaftspädagogik, 21, 111–130.

16.

Ding

Bosker

R. J.

Harskamp

E. G.

(2011). Exploring gender and gender pairing in the knowledge elaboration processes of students using computer-supported collaborative learning. Computers & Education, 56(2), 325–336.

17.

Erkens

Bodemer

(2019). Improving collaborative learning: Guiding knowledge exchange through the provision of information about learning partners and learning contents. Computers & Education, 128, 452–472.

18.

Espasa

Guasch

Mayordomo

R. M.

Martinez-Melo

Carless

(2018). A Dialogic Feedback Index measuring key aspects of feedback processes in online learning environments. Higher Education Research & Development, 37(3), 499–513.

19.

Faul

Erdfelder

Lang

A. G.

Buchner

(2007). G* Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. https://doi.org/10.3758/BF03193146

20.

Fischer

Mandl

(2023). Facilitating knowledge convergence in videoconferencing environments: The role of external representation tools. In G. Stahl (Ed.), Computer support for collaborative learning (pp. 623–624). Routledge.

21.

Gombert

Fink

Giorgashvili

Jivet

Di Mitri

Yau

Frey

Drachsler

(2024). From the automated assessment of student essay content to highly informative feedback: A case study. International Journal of Artificial Intelligence in Education, 34, 1378–1416. https://doi.org/10.1007/s40593-023-00387-6

22.

Guasch

Espasa

Martinez-Melo

(2019). The art of questioning in online learning environments: The potentialities of feedback in writing. Assessment & Evaluation in Higher Education, 44(1), 111–123.

23.

Hattie

Timperley

(2007). The power of feedback. Review of Educational Research, 77(1), 81–112.

24.

Hernández-Sellés

Muñoz-Carril

P. C.

González-Sanmamed

(2020). Interaction in computer supported collaborative learning: An analysis of the implementation phase. International Journal of Educational Technology in Higher Education, 17(1), 1–13.

25.

Hsieh

Y. H.

Tsai

C. C.

(2012). The effect of moderator’s facilitative strategies on online synchronous discussions. Computers in Human Behavior, 28(5), 1708–1716.

26.

Huang

Han

Jiang

(2023). Examining the relationship between peer feedback classified by deep learning and online learning burnout. Computers & Education, 207, 104910.

27.

Huang

Mao

Zhang

(2024). The influence of artificial intelligence technology on college students’ learning effectiveness from the perspective of constructivism—taking ChatGPT as an example. Journal of Education, Humanities and Social Sciences, 30, 40–46.

28.

Hwang

G.-J.

Chang

S. C.

(2021). Facilitating knowledge construction in mobile learning contexts: A bi-directional peer-assessment approach. British Journal of Educational Technology, 52(1), 337–357. https://doi.org/10.1111/bjet.13001

29.

Ifenthaler

Schumacher

Sahin

(2021). System-based or teacher-based learning analytics feedback—what works best? [Conference session]. 2021 International Conference on Advanced Learning Technologies (ICALT) (pp. 184–186). Tartu, Estonia: IEEE. https://doi:10.1109/ICALT52272.2021.00062

30.

Ito

Umemoto

(2022). Self-regulation, co-regulation, and socially shared regulation of motivation for collaborative activity: Comparison between university students and working adults. Japanese Psychological Research, 64(4), 397–409.

31.

Järvelä

Hadwin

A. F.

(2013). New frontiers: Regulating learning in CSCL. Educational Psychologist, 48(1), 25–39.

32.

Jeong

Chi

M. T.

(2007). Knowledge convergence and collaborative learning. Instructional Science, 35, 287–315.

33.

Jeong

Hmelo-Silver

C. E.

(2016). Seven affordances of computer-supported collaborative learning: How to support collaborative learning? How can technologies help?. Educational Psychologist, 51(2), 247–265.

34.

Pan

Cambria

Marttinen

Philip

S. Y.

(2021). A survey on knowledge graphs: Representation, acquisition, and applications. IEEE Transactions on Neural Networks and Learning Systems, 33(2), 494–514.

35.

Jiang

J. P.

J. Y.

Zhang

Y. B.

Yin

X. C.

(2023). Fostering college students’ critical thinking skills through peer assessment in the knowledge building community. Interactive Learning Environments, 31(10), 6480–6496.

36.

Jin

Lim

Y. K.

(2019). Effects of peer feedback types and feedback acceptance levels on academic achievement in middle school project-based learning. Educational Technology International, 20(1), 57–81.

37.

Kalyuga

(2009). Knowledge elaboration: A cognitive load perspective. Learning and Instruction, 19(5), 402–410.

38.

Kapur

Voiklis

Kinzer

C. K.

(2011). A complexity-grounded model for the emergence of convergence in CSCL groups. In Puntambekar

Erkens

Hmelo-Silver

(Eds.), Analyzing interactions in CSCL: Methods, approaches and issues (pp. 3–23). Springer.

39.

Kerman

N. T.

Noroozi

Banihashem

S. K.

Karami

Biemans

H. J.

(2024). Online peer feedback patterns of success and failure in argumentative essay writing. Interactive Learning Environments, 32(2), 614–626.

40.

Kielstra

Molenaar

van Steensel

Verhoeven

(2022). Supporting socially shared regulation during collaborative task-oriented reading. International Journal of Computer-Supported Collaborative Learning, 17(1), 65–105. https://doi.org/10.1007/s11412-022-09365-x

41.

Kwon

(2020). Student-generated awareness information in a group awareness tool: What does it reveal?. Educational Technology Research and Development, 68(3), 1301–1327.

42.

Lai

C.-Y.

Chen

L.-J.

Yen

Y.-C.

Lin

K.-Y.

(2020). Impact of video annotation on undergraduate nursing students’ communication performance and commenting behaviour during an online peer-assessment activity. Australasian Journal of Educational Technology, 36(2), 71–88.

43.

Lakens

(2013). Calculating and reporting effect sizes to facilitate cumulative science: A practical primer for t-tests and ANOVAs. Frontiers in Psychology, 4, 863.

44.

Latifi

Noroozi

Talaee

(2021). Peer feedback or peer feedforward? Enhancing students’ argumentative peer learning processes and outcomes. British Journal of Educational Technology, 52(2), 768–784.

45.

Lund

(2019). Building and regulating cognitive, linguistic, and interactional aspects of knowledge between the individual and the group. Computers in Human Behavior, 100, 370–383.

46.

Martha

A. S. D.

Santoso

H. B.

Junus

Suhartanto

(2023). The effect of the integration of metacognitive and motivation scaffolding through a pedagogical agent on self- and co-regulation learning. IEEE Transactions on Learning Technologies, 16(4), 573–584. 10.1109/TLT.2023.3266439

47.

Mayweg-Paus

Zimmermann

N. T.

Pinkwart

(2021). A review of technologies for collaborative online information seeking: On the contribution of collaborative argumentation. Education and Information Technologies, 26, 2053–2089.

48.

Meijer

Hoekstra

Brouwer

Strijbos

J. W.

(2020). Unfolding collaborative learning assessment literacy: A reflection on current assessment methods in higher education. Assessment & Evaluation in Higher Education, 45(8), 1222–1240.

49.

Mercier

E. M.

(2017). The influence of achievement goals on collaborative interactions and knowledge convergence. Learning and Instruction, 50, 31–43.

50.

Pietarinen

Palonen

Vauras

(2021). Guidance in computer-supported collaborative inquiry learning: Capturing aspects of affect and teacher support in science classrooms. International Journal of Computer-Supported Collaborative Learning, 16, 261–287.

51.

Quera

Bakeman

Gnisci

(2007). Observer agreement for event sequences: Methods and software for sequence alignment and reliability estimates. Behavior Research Methods, 39(1), 39–49.

52.

Qureshi

M. A.

Khaskheli

Qureshi

J. A.

Raza

S. A.

Yousufi

S. Q.

(2023). Factors affecting students’ learning performance through collaborative learning and engagement. Interactive Learning Environments, 31(4), 2371–2391.

53.

Resta

Laferrière

(2007). Technology in support of collaborative learning. Educational Psychology Review, 19, 65–83.

54.

Roschelle

(1996). Learning by collaborating: Convergent conceptual change. In Koschmann

(Ed.), CSCL: Theory and practice of an emerging paradigm (pp. 209–248). Lawrence Erlbaum Associates Inc. http://www.jstor.org/stable/1466609

55.

Roschelle

Teasley

(1995). The construction of shared knowledge in collaborative problem solving. In O’Malley

(Ed.), Computer-supported collaborative learning (pp. 69–97). Springer.

56.

Sadita

Hirashima

Hayashi

Furtado

P. G.

Junus

Santoso

H. B.

(2020). The effect of differences in group composition on knowledge transfer, group achievement, and learners’ affective responses during reciprocal concept mapping with the Kit-Build Approach. Research and Practice in Technology Enhanced Learning, 15(1), 1–19.

57.

Sedrakyan

Malmberg

Verbert

Järvelä

Kirschner

P. A.

(2020). Linking learning behavior analytics and learning science concepts: Designing a learning analytics dashboard for feedback to support learning regulation. Computers in Human Behavior, 107, 105512.

58.

Sedrakyan

Mannens

Verbert

(2019). Guiding the choice of learning dashboard visualizations: Linking dashboard design and data visualization concepts. Journal of Computer Languages, 50, 19–38.

59.

Shao

Cheng

Wang

(2024). How peer feedback with regulation scripts contributes to the development of critical thinking in dialogues: Strengthening cognitive and affective feedback content. Interactive Learning Environments, 32(10), 6121–6140.

60.

Shin

Kim

Song

(2020). Types and timing of scaffolding to promote meaningful peer interaction and increase learning performance in computer-supported collaborative learning environments. Journal of Educational Computing Research, 58(3), 640–661.

61.

Singh

(2021). Quasi experimental design in science psychology. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3793568

62.

Souza

P. V.

Pereira

M. M.

Lourenço

A. B.

(2024). Hierarchical organization in concept maps as a path to explain the elaboration of knowledge in the history of science. Science Education International, 35(1), 21–29. https://doi.org/10.33828/sei.v35.i1.3

63.

Stahl

Koschmann

Suthers

(2014). Computer-supported collaborative learning. In Sawyer

R. K.

(Ed.), The Cambridge handbook of the learning sciences (pp. 479–500). Cambridge University Press.

64.

Stegmann

Wecker

Weinberger

Fischer

(2012). Collaborative argumentation and cognitive elaboration in a computer-supported collaborative learning environment. Instructional Science, 40(2), 297–323.

65.

Liang

J. C.

Zheng

Tsai

C. C.

(2023). Co-regulation strategies and their associations with writing self-efficacy in a computer-mediated collaborative writing setting. Journal of Second Language Writing, 59, 100972. https://doi.org/10.1016/j.jslw.2023.100972

66.

Sun

Looi

C. K.

Yang

Jia

(2025). Exploring students’ learning performance in computer-supported collaborative learning environment during and after pandemic: Cognition and interaction. British Journal of Educational Technology, 56(1), 128–149. https://doi.org/10.1111/bjet.13492

67.

Tan

De Weerd

J. G.

Stoyanov

(2021). Supporting interdisciplinary collaborative concept mapping with individual preparation phase. Educational Technology Research and Development, 69, 607–626. https://doi.org/10.1007/s11423-021-09963-w

68.

Tan

J. S.

Chen

(2022). Peer feedback to support collaborative knowledge improvement: What kind of feedback feed-forward?. Computers & Education, 187, 104467.

69.

Tempelaar

D. T.

Rienties

Giesbers

(2015). In search for the most informative data for feedback generation: Learning analytics in a data-rich context. Computers in Human Behavior, 47, 157–167.

70.

Theobald

Bellhäuser

(2022). How am I going and where to next? Elaborated online feedback improves university students’ self-regulated learning and performance. The Internet and Higher Education, 55, 100872.

71.

Thiel de Gafenco

Weinert

Janson

Klusmeyer

Leimeister

J. M.

(2024). Shared digital artifacts—Co-creators as beneficiaries in microlearning development. Education and Information Technologies, 29(6), 7129–7154. https://doi.org/10.1007/s10639-023-12074-z

72.

Tseng

S. C.

Tsai

C. C.

(2007). On-line peer assessment and the role of the peer feedback: A study of high school computer course. Computers & Education, 49(4), 1161–1174.

73.

Tseng

S.-S.

(2024). Maximizing the impact of dialogic peer feedback on student engagement: The role of regulation support. Educational Technology & Society, 27(2), 133–148. https://doi.org/10.30191/ETS.202404_27(2).RP04

74.

Ustun

A. B.

Zhang

Karaoğlan-Yilmaz

F. G.

Yilmaz

(2023). Learning analytics-based feedback and recommendations in flipped classrooms: An experimental study in higher education. Journal of Research on Technology in Education, 55(5), 841–857.

75.

Wang

S. L.

Hong

H. T.

(2018). The roles of collective task value and collaborative behaviors in collaborative performance through collaborative creation in CSCL. Educational Technology Research and Development, 66, 937–953.

76.

Weinstein

C. E.

Mayer

R. E.

(1986). The teaching of learning strategies. In Wittrock

C. M.

(Ed.), Handbook of research in teaching (pp. 315–327). Macmillan.

77.

Weldon

Weingart

L. R.

(1993). Group goals and group performance. British Journal of Social Psychology, 32(4), 307–334.

78.

Wen

Song

(2021). Learning analytics for collaborative language learning in classrooms: From the holistic perspective of learning analytics, learning design and teacher inquiry. Educational Technology & Society, 24(1), 1–15.

79.

Yildiz Durak

. (2024). Feedforward- or feedback-based group regulation guidance in collaborative groups. Journal of Computer Assisted Learning, 40(2), 410–436.

80.

Zabolotna

Malmberg

Järvenoja

(2023). Examining the interplay of knowledge construction and group-level regulation in a computer-supported collaborative learning physics task. Computers in Human Behavior, 138, 107494.

81.

Zambrano

R J.

Kirschner

Sweller

Kirschner

P. A

. (2023). Effect of task-based group experience on collaborative learning: Exploring the transaction activities. British Journal of Educational Psychology, 93(4), 879–902.

82.

Zhang

Costley

Courtney

Shulgina

Fanguy

(2024). The impact of different peer feedback types on student academic writing performance from dyadic and individual analyses. Education and Information Technologies, 30(5), 6339–6366. https://doi.org/10.1007/s10639-024-13032-z

83.

Zheng

(2017). Knowledge building and regulation in computer-supported collaborative learning. Springer.

84.

Zhu

Ergulec

(2023). A review of collaborative assessment strategies in online learning. Distance Education, 44(3), 522–543. https://doi.org/10.1080/01587919.2022.2150127

85.

Zheng

Niu

Long

Fan

(2023). An automatic knowledge graph construction approach to promoting collaborative knowledge building, group performance, social interaction and socially shared regulation in CSCL. British Journal of Educational Technology, 54(3), 686–711. https://doi.org/10.1111/bjet.13283